Identification of research hypotheses and new knowledge from scientific literature

Table 3 Statistics comparing our versions of the GENIA-MK and EU-ADR corpora, both annotated with new knowledge and research hypothesis labels

	GENIA-MK	EU-ADR
Base type for annotations	Events	Relations
Number of annotations	6899	622
Number of abstracts	150	159
Number of new knowledge annotations	2356 (34.2%)	406 (65.3%)
Number of research hypothesis annotations	366 (5.31%)	38 (6.11%)

The GENIA-MK corpus is much more densely annotated than the EU-ADR corpus, with over ten times more annotated events in the former than annotated relations in the latter. Research Hypotheses are particularly sparse in both corpora, constituting just over 5% of all annotated relations and events in each case. There is a disparity in the proportion of New Knowledge between the two corpora, in part because the EU-ADR corpus appeared to favour the annotation of relationships denoting New Knowledge

ISSN: 1472-6947